Experiment 1
Measurements and Uncertainties
The density of a solid metal cylinder is found from measurements made of its mass and
volume. A double pan balance is used to find the mass, and a meter stick and a vernier caliper are
alternately used to measure the height and diameter of the cylinder in order to emphasize the
relative differences in the uncertainties of the two measuring instruments. These uncertainties are
shown to propagate through the calculation resulting in an uncertainty associated with the
determination of density.
Theory
The density of a homogeneous material is defined as the mass per unit volume, or
=
mass
m
=
volume v
(1)
The units associated with the density are kg/m3, g/cm3, and slug/ft3.
Whenever measurements are performed, two factors contribute to the total uncertainty
associated with those measurements. First, an uncertainty exists which is associated with the
instrument itself and its construction. This uncertainty is generally included in the specifications of
the instrument and expressed as a plus or minus value or as a percentage. For example, Figure 1
shows the measurement to the right edge of a book to be 25.74 cm. (The edge of the book is
between 25.7 and 25.8 cm, and by estimating where the edge lies between these two positions, one
gets 25.74 cm. The 4 is an estimated value that can be different for each observer.) Because an
uncertainty exists due to the limitation of the measuring instrument, the measurement made with
the meter stick might be expressed as 25.74  0.02 cm, or if a percentage uncertainty such as 0.1%
is given, then the uncertainty in the reading is 0.1% times 25.74 or  0.03 (0.02574) cm. The
uncertainty described above is the instrumental uncertainty.
1
The second uncertainty is an uncertainty
associated with the ability of the experimenter to
perform the measurement accurately due to the
characteristics of the object to be measured. For
example, if the edge of the book is damaged and
not sharply defined, the experimenter must estimate
the position of the edge and this estimate introduces
an additional uncertainty that must be evaluated
subjectively by the experimenter. Note that this
uncertainty must be assigned by the experimenter.
This is referred to as the experimental uncertainty.
(Other methods of error analysis that require
multiple measurements result in a more objective
evaluation of the uncertainty.)
Figure 1. A measurement of the length of the book
gives a value of 25.74 cm.
When the instrumental uncertainty is combined with the experimenter's estimated
uncertainty, the result is the total uncertainty associated with the measurement.
This type of uncertainty is a form of random error. Random errors can be minimized (and
always should be) but cannot be entirely eliminated, and they are equally likely to be too high or too
low (hence “random”). The only recourse available to the experimenter in this case is to assess the
uncertainty as carefully as possible and to report it honestly. This is the point of this experiment!
The other major category of experimental error is systematic error. Systematic errors may
be caused by faulty equipment (ie. failure to zero a balance) or because incorrect, simplifying
assumptions have been made about the experiment (ie. that a surface is frictionless). Unlike random
errors, systematic errors are biased in one direction (always too high or too low). Also, unlike
random errors, systematic errors can be eliminated (and always should be) by repairing the
equipment or by correcting the assumptions about the experiment. A proper experiment should
never have systematic errors. We will learn more about this kind of error in subsequent
experiments.
Propagation of Errors
Two methods are shown that demonstrate how the uncertainties associated with
measurements are propagated through the calculations to result in an uncertainty in the final result.
Suppose a physical quantity A is to be found where
A=C
X 2Y
Z
(3)
C is assumed to be a constant and X, Y, and Z are measured quantities with corresponding
uncertainties  X ,Y , and  Z . We assume that these are the total random uncertainties.
2
The first method requires the substitution of the uncertainties into the expression for A. That
is,
A  A  C
( X  X )2(Y  Y )
.
( Z  Z )
The two terms, A+ A and A- A represent the maximum and minimum values A can have
when the uncertainties are taken into account. The maximum value, A+ A , is found by choosing
either the plus or minus sign preceding the uncertainties so as to maximize A. In this case, the plus
signs are chosen for the terms in the numerator and the minus sign in the denominator. The
maximum value of A is then
( X  X )2(Y  Y )
A+ A  C
(4)
( Z  Z )
In a similar fashion, the minimum value of A is
A- A  C
( X  X )2(Y  Y )
( Z  Z )
(5)
The value of A is calculated from (3), and A is found either by subtracting (3) from (4) or (5).
The second method of finding how uncertainties are propagated utilizes differentials.
Recall that differentials represent infinitesimal changes, and that small finite changes are
approximately equal to the infinitesimal changes. That is,
dQ  Q
as long as Q is small.
To find the uncertainty A, the differential dA is taken, then all the differentials are replaced
by their appropriate small finite changes. In order to simplify the taking of the differential, the
logarithm is taken first, then the differential. This produces the same result. First, take the natural
log of A,
lnA = lnC + 2lnX + lnY – lnZ.
Now take the differential of this expression,
dA dC
dX dY dZ

2


A
C
X
Y
Z
Multiply both sides of this equation by A and recall that the term dC/C is zero because C is
constant, then
 dX dY dZ 


dA = A 2
Y
Z 
 X
3
The infinitesimal uncertainties are replaced with the corresponding small finite uncertainties
 X ,Y , and  Z (the values of X , Y , andZ are positive), then
      

 


 
A  A  2
(6)
In the above expression, the plus or minus sign associated with
order to maximize .
,  , and  is chosen in
Now that A from (3) and . from (6) are found, the range within which the experimental
value of A should fall is defined. This range is from    to    . The range can be
indicated on a one-dimensional graph using error bars. (See Figure 2.)
The meaning of this range is that the "true"
value of A must lie within the range + . provided
the uncertainties X , Y , andZ have been correctly
estimated and the measurements and calculations have
been performed correctly. If the "true" or best known
value of A lies outside the range, then this indicates that
a measurement or computational error may have been
made and/or the uncertainties in the measured values of
X, Y, and Z were not correctly estimated. Anytime this
occurs, the experimenter should reexamine the
calculations, measurements, and the estimations of
uncertainties.
Figure 2. A one-dimensional graph showing
the range of the experimental value
of A.
Apparatus
o meter stick,  0.05 cm
o vernier caliper,  0.005cm
o double pan balance,  0.2 gm
o solid metal cylinder
Procedure
1) Find the mass of the solid cylinder, m, using the double pan balance. Record m and its
uncertainty m . Remember that the uncertainty consists of two parts: the instrumental
uncertainty and the uncertainty associated with the experimenter's ability to accurately
perform the measurement.
4
2) Use the meter stick to measure the height, h, and the diameter, D, of the cylinder.
Record h and D and their uncertainties h and D . Again remember that the
uncertainties consist of two parts.
3) Use the vernier caliper to measure the height, h, and the diameter, D, of the cylinder.
Record h and D and their uncertainties, h and D .
Analysis
The density of a sample of material is given by (1), that is,
m
=
v
(1)
Where m is the mass of the sample and V is the volume. The volume of a cylinder is
V=
D 2h
4
,
(7)
where D is the diameter and h is the height of the cylinder. The density can then be written as
=
4m
,
D 2h
(8)
and the uncertainty in the density as (using the second method described earlier)
D h 
 m
2
 .
D
h
m
   
(9)
Use (8) and (9) and the meter stick measurements to calculate the density of the sample and
its uncertainty. Graph these results on a one-dimensional error graph similar to Figure 2. Label the
error bar as the "meter stick determination" of the density.
Use (8) and (9) and the vernier caliper measurements to calculate the density and its
uncertainty. Graph these results on the same one-dimensional graph described above with the error
bar displaced vertically above the previous results. Label the error bar as the "vernier caliper
determination" of the density.
Ask the instructor for the type of metal from which the cylinder is made, and look up the
theoretical value of the density from your text, the Handbook of Chemistry and Physics, or from
some other source. Plot this value on the same graph described above but displace it vertically
above the previous results. Label this value as the "best known value."
Report the experimental values of the density, their corresponding uncertainties, and the
book value of the density in a table of results.
5
Conclusions
Look carefully at your results. Were the random errors estimated adequate to account for the
difference between your obtained results and the known density values? In other words, did your
experimental result fall within the expected “uncertainty range”?
If your results were outside the expected uncertainty range, describe the major sources of
systematic error that could account for the differences. Explain how the sources of error would have
affected the experimental value.
Extensions & Questions
1) The basic “rule of significant digits” states that the total number of significant digits in a
calculated quantity is equal to the least number of significant digits in the measured
quantities that are used to calculate a result. Did your results reflect this simple rule?
2) Did you observe evidence of a systematic error in your results, with either the ruler of
the calipher-derived density?
3) How would the little hole in the cylinder affect the results? Would this be a systematic
or a random error? Did you see evidence for this type of error in your results?
6